The Grid Computing and Distributed Systems (GRIDS) Laboratory at the
University of Melbourne is expanding its Grid computing work with the
new Gridbus project.
The open source Gridbus project, at www.gridbus.org, is focused on
developing technology that enables Grid computing and business, hence
the name Gridbus. The Gridbus project team is developing cluster and
Grid technologies (middleware, tools and applications) that deliver
end-to-end quality of services depending on user requirements, according
to program leader Rajkumar Buyya.
The technologies include Economic Grid Scheduler, Cluster Scheduler
(Libra), Grid modeling and simulation (GridSim), Data Grid broker,
GridBank, and GUI tools for workflow management and composition of
distributed applications from legacy software components. The Gridbus
scheduling system aggregates or leases of services of distributed
resources depending on their availability, capability, performance,
cost, and users quality-of-service requirements. The Gridbus technology
development is driven by requirements of various applications: Drug
Design, High Energy Physics, and Brain Activity Analysis. The World Wide
Grid (WWG) testbed used in the research contains resources from
organizations around the globe, Buyya said.
"While our previous work (www.buyya.com/thesis) on economic-based
resource management and scheduling for Grid computing has been very
successful, it was limited to a single parallel programming paradigm and
a commodity economic model with price of each resource assumed to remain
the same for the duration of application processing," said Buyya.
"It assumes that the application tasks are compute-intensive and
independent of each other," he said. "Although this is a dominant model
for a class of applications that are being explored, it cannot be
applied directly for applications that are: data intensive; tasks that
have some interdependency; and tasks that may need to communicate
amongst themselves. Also, there exist many other applications that need
support for different parallel programming models. There exist several
other economic models such as contract-net and auctions - each have the
potential to serve as an effective means for managing resources under
different scenarios."
The key objective of the Gridbus project is to build on previous
contributions and explore the development of service-oriented
architecture and high-level scheduling services for different
application programming paradigms and economic models, Buyya said. The
programming framework aims to unify both computational and data
intensive application requirements, and the distributed computing
runtime machinery supports secure and transparent services to aggregate
and unify emerging low-level Grid, P2P, and Web services, he said.
The current local resource management systems do not provide any
guaranteed quality of services, he said. "To overcome this limitation,
we propose to develop an economic-scheduler for clusters and enable the
development of high-level services delivering hard, instead of soft,
quality-of-services."
The Gridbus project extends previous work on Grid economy and scheduling
to support different application models, different economy models, data
models, and architecture models - both Grids and P2P networks, Buyya
said. At the resource level, it supports QoS driven resource scheduler
(e.g., economy driven cluster scheduler), which helps to enforce
allocation of resources explicitly.
Other Gridbus initiatives include: a GridBank mechanism that supports a
secure Grid-wide accounting and payment handling to enable both
cooperative and competitive economy models for resource sharing; the
GridSim simulator, which is being extended to support simulation of
these concepts for performance evaluation; GUI tools for enabling
distributed processing of legacy applications; and applying the work to
various application domains (high-energy physics, brain activity
analysis, drug discovery, data mining, GridEmail, automated management
of e-commerce).
Events
IE 7
Firefox 2.0
